Sei sulla pagina 1di 18

THE CASIMIR EFFECT

Joseph Cugnon
University of Li`ege, AGO Department, allee du 6 Ao ut 17, b at. B5,
B-4000 Li`ege 1, Belgium
Abstract
The Casimir eect is usually interpreted as due to the modication of the zero point
energy of QED when two perfectly conducting plates are put very close to each other,
and, consequently, as a proof of the reality of this zero point energy. The Dark
Energy, necessary to explain the acceleration of the expansion of the Universe is
sometimes viewed as another proof of the same reality. The usual interpretation of
the Casimir eect is however challenged by some authors who rather consider it as
a giant van der Waals eect. All these aspects are discussed.
1 Introduction
The Casimir eect corresponds to the force acting beetween two uncharged parallel condensor plates.
It is customarily attributed to the change in zero point energy of the electromagnetic vacuum ex-
tending between the plates with respect to the one of the vacuum contained in the same region in the
absence of plates. The zero point energy is supposed to result from the standard quantization of the
free electromagnetic eld. This energy is not directly observable, but the force between the two plates
results from the change of the zero point energy contained between the plates when the latter are
moved apart from each other. The Casimir eect is generally considered as a proof of the reality
of the zero point energy. The dark energy, seemingly necessary to explain the observed accelerating
expansion of the Universe is sometimes advocated as another proof of the same reality. All these
aspects are shortly examined below.
2 The usual derivation of the Casimir eect
Let us consider an ideal condensor with innite perfectly conducting plates in the x and y directions
separated by a distance d along the z direction as shown in Fig. 1.
The interaction energy between the two plates can be dened as the dierence between the zero
point energies contained in the space between the planes, in the two respective congurations. This
Preprint submitted to Elsevier Science 13 July 2009
supposes that the eld outside the cavity is not changed, which, of course, holds classically. For a
discussion of this point, as well as for corrections due the niteness of the plates, see Ref. [1]. In
general, the zero point energy of the electromagnetic eld is given
E
cav
=

k
g
s
(k)

k
2
, (1)
where the sum runs over the normal modes of the eld, where the
k
s are the frequencies of these
modes and where g
s
(k) is the degeneracy of the mode k , due to polarization.
d
z 0
Fig. 1. I deal condensor with innite extension in the x and y directions. The origin of the z axis is located
on the left plate.
Let us start with the case of the cavity. In order to identify the modes more easily, let us consider,
as asual, eld congurations which are periodic in the x and y directions with periods L
x
and L
y
,
respectively. The normal modes are determined by the boundary conditions on the surface of the
conductors. We remind that the tangential electric eld and the normal component of the magnetic
eld should vanish on these boundaries. These conditions are realized when the vector potential

is
given, for k
z
,= 0, by

= e
ikxx
e
ikyy
sink
z
z e
i
k
t
, (2)
with
k
x
= n
x
2
L
x
, k
y
= n
y
2
L
y
, (3)
2
where n
x
and n
y
are integer numbers and with
k
z
= n
z

d
, (4)
where n
z
is a positive integer, the vector is the polarisation vector and where

k = (k
x
, k
y
, k
z
). It
is perpendicular to the vector

k : .

k = 0, as explained below. The quantity



can be viewed as the
vector potential in the Coulomb gauge. The electric eld is then proportional to the vector potential:

E = i
k
e
ikxx
e
ikyy
sink
z
z e
i
k
t
. (5)
where is the polarization vector. The magnetic eld can be written as

B =

k e
ikxx
e
ikyy
cosk
z
z e
i
k
t
. (6)
The boundary conditions are satised as follows. The vanishing of the tangential electric eld is
guaranteed by the presence of the sine function in

and the values of k
z
. The vanishing of the
normal magnetic eld requires
e
z
.(

k ) = 0, (7)
which should hold together with

k. = 0, (8)
the latter relation resulting from

.

E=0. The last two equations are explicitated as:


k
x

y
k
y

x
= 0, (9)
k
x

x
+ k
y

y
+ k
z

z
= 0. (10)
For each value of

k with k
z
,= 0, there is an innity of solutions and it is thus possible to select
two modes corresponding to two vectors satisfying the last two equations and perpendicular to
each other. The conditions are satised dierently for the modes

k with k
z
= 0. Form (2) is not
satisfactory, since is then vanishing identically. However, the form

= e
ikxx
e
ikyy
e
i
k
t
, (11)
where sin(k
z
z) has been replaced by cos(k
z
z) (equal to unity), is also a solution of the Laplace
equation in this case. The electric and magnetic elds are given by

E = i
k
e
ikxx
e
ikyy
e
i
k
t
. (12)
3
and

B =

k e
ikxx
e
ikyy
e
i
k
t
. (13)
The vanishing of the tangential electric eld can now only be guaranteed by a normal vector, which
also guarantees the vanishing of the normal component of the magnetic eld (13), since the vector

k
has only tangential component. However, the vector does still have to fulll Eqs. (9,10). For k
z
= 0,
the only solution is = e
z
: for these modes, there is only one possible polarisation (g
s
=1). These
k
z
= 0 modes are often forgotten in the literature (see for instance Ref. [2]) , leading to confusing
statements about regularisation procedures.
The energy of the cavity for the k
z
,= 0 modes (indicated by the prime) is given by
E

cav
= c
+

nx=
+

ny=

nz=1
_
_
n
x
2
L
x
_
2
+
_
n
y
2
L
x
_
2
+
_
n
z

d
_
2
_
1/2
. (14)
As usual, one replaces the summation on n
x
and n
y
by integration on continuous variables:
E

cav
= cL
x
L
y
+
_

dn
x
+
_

dn
y

nz=1
_
_
n
x
2
L
x
_
2
+
_
n
y
2
L
x
_
2
+
_
n
z

d
_
2
_
1/2
, (15)
which causes no dierence when the limit L
x
, L
y
is taken. Changing variables from n
x
, n
y
to
k
x
, k
y
, dividing the expression by L
x
L
y
and taking the limit of L
x
and L
y
tending to innity, yields
for the energy per unit area:
E

cav
S
= c

n=1
+
_

dk
x
2
+
_

dk
y
2
_
k
2
x
+ k
2
y
+
_
n
d
_
2
_
1/2
. (16)
Finally, one integrates over the angle of the wave vector in the x y plane, introduces the auxiliary
variable u = ((k
2
x
+ k
2
y
)d
2
)/
2
= k
2

d
2
/
2
and obtains
E

cav
S
=
c
2
4d
3

n=1

_
0
du
_
u + n
2
_
1/2
. (17)
We still have to add the contribution of the k
z
= 0 (or n = 0) modes. One has:
E
cav
S
=
c
2
4d
3
_
_

n=1

_
0
du
_
u + n
2
_
1/2
+
1
2

_
0
du

u
_
_
. (18)
This expression is divergent, as is the similar expression for the energy of the free eld. It may be
hoped that the dierence between the two expressions is nite.
4
We need the value of the energy of the free eld in the volume of the cavity. In general, the energy
per unit volume is given by
1
E
free
V
= c
_
d
3

k
(2)
3
k. (19)
It is advantageous to rewrite this expression as
E
free
V
=
c
(2)
3
_
d
2

_
dk
z
_
k
2

+ k
2
z
, (20)
or as
E
free
V
=
c
(2)
2

_
0
k

dk

+
_

dk
z
_
k
2

+ k
2
z
. (21)
Taking account of the fact that the integrand is an even function of k
z
and introducing the auxiliary
variables u = k
2

d
2
/
2
and x = k
z
/d lead to
E
free
V
=
c
2
4d
4

_
0
dx

_
0
du

u + x
2
. (22)
The energy of the free eld in the volume of the cavity is thus obtained by multiplying this expression
by the volume V (equal to Sd). One then obtains for the change in the zero point energy per unit
surface:
E
S
=
E
cav
S

E
free
S
=
c
2
4d
3

1
2

_
0
du

u +

n=1

_
0
du
_
u + n
2
_
1/2

_
0
dx

_
0
du
_
u + x
2
_
1/2
. (23)
All terms in the rhs are divergent. We will come to this problem soon. It is remarkable that this
expression involves the dierence between the integral from zero to innity of the function f(x) =
_

0
du (u + x
2
)
1/2
and the sum of the values of this function on the positive integers. There is a famous
theorem by Euler and McLaurin connecting these quantities, in general. It states that:

n=1
f(n) =

_
0
f(x)dx
1
2
[f(0) + f()] +
1
12
[f

() f

(0)]
1
720
[f

() f

(0)] + (24)
where the dots indicates similar terms for higher order odd derivatives. The coecients in front of
the brackets are related to the Bernoulli numbers B
i
: they are equal to B
2k
/(2k)! for the term
involving the derivatives of order 2k 1. See Ref. [3].
1
The

k = 0 mode does not pose any worry, since its contribution is vanishing.
5
We can write the function f(x) mentioned above as
f(x) =

_
0
du
_
u + x
2
_
1/2
=

_
x
2
dt

t. (25)
Formally, considering the dependence upon x through the lower bound of the integral only, one has:
f

(x) = 2x
2
, f

(x) = 4x, f

(x) = 4 (26)
and all higher derivatives are vanishing. So, retaining the single nondivergent term (which corresponds
to f

(0)), one nally obtains:


E
S
=
c
2
720d
3
. (27)
This is the expression of the Casimir eect, which looks universal and which depends only upon the
two fundamental constants and c and on the distance d.
Let us comment on the divergence problems rst. Of course, expression (23) and the Euler-MacLaurin
theorem apply when the quantities are convergent. However, one can make the nal result meaningful
by regularizing the integral and the sum. The regularisation at innity is not a problem. It is easy to
introduce a suitable integration factor. For instance, it is easy to see that all terms at innity vanish
owing to the substitution
f(x)

_
0
du
_
u + x
2
_
1/2
e
(u+x
2
)
, (28)
The terms corresponding to higher order derivatives (at x = 0 as well as at x = ) are nite and
vanish as 0. Similarly, the rst and third derivatives at x = 0 are incremented by quantities
that vanishes as 0. Actually, the regularisation can be achieved by any cut-o function which
decreases suciently rapidly at large x (non necessarily exponentially), which goes to unity as x goes
to zero with vanishing derivatives, like the functions g(x) = 1 exp(a/x).
3 Which force?
The force (per unit surface) acting between the plates is given by
F
S
=
c
2
240d
4
. (29)
6
The negative sign corresponds to an attractive force. It is a tiny force. For instance, at d=1 m, it
amounts to F/S= 410
4
N/m
2
. Of course, due to the fourth power, it increases very rapidly as the
distance decreases. At d=1 nm, the force reaches F/S= 410
8
N/m
2
.
Needless to say that the experimental verication of the Casimir eect has taken quite a long time.
Among the unsuccessful trials, one should mention the experiment by Sparnaay [4], which although
unsuccessful, has nevertheless identied the main diculties: a perfect parallelism of the plates, a
lack of impurities (which may scatter the normal modes) and the elimination of the residual charges.
Let us also mention the experiment of Derjaguin et al [5], who were the rst to obtain a meaningful
result, verifying the predictions at the 60% level, before the experiments by Lamoreaux [6] and Ederth
[7] who veried the theoretical value with an accuracy of 1%.
4 The Casimir force and the van der Waals eect
4.1 Introduction
The Casimir force may be viewed as a quantum interaction between two neutral objects. Of course,
the conducting properties of these objects should be taken into account at some point. But for the
moment, let us consider the Casimir force as the force acting between macroscopic neutral objects
and address the question whether there is some relationship with the force acting in another system
of this kind, namely the system of two neutral atoms. We will examine this question in a bit historical
perspective, which helps to understand the relationship.
4.2 The van der Waals force in the simplest approach
The van der Waals interaction has been calculated microscopically for the rst time by London [8].
The hamiltonian of the system can be written as
H =H
1
+ H
2
+ H

,
H

=
Z
1
Z
2
e
2
[

R
1


R
2
[

Z
1

i=1
Z
2
e
2
[

R
1
+ r
i


R
2
[

Z
2

j=1
Z
1
e
2
[

R
2
+ r
j


R
1
[
+
Z
1

i=1
Z
2

j=1
e
2
[

R
2
+ r
j


R
1
r
i
[
. (30)
In this equation, Z
1
and Z
2
are the charge numbers of the nuclei and r
i
, r
j
are the coordinates of the
electrons with respect to the position of the respective nuclei. Expanding H

up to second order in
the electron coordinates r
i
, r
j
(which is presumably sucient if the distance r = [

R
1


R
2
[ is large
in comparison with the atomic sizes), one has
H

= 2
Z
2
e
[

R
1


R
2
[
3
(

R
1


R
2
).
Z
1

i=1
e r
i
2
Z
1
e
[

R
1


R
2
[
3
(

R
2


R
1
).
Z
2

j=1
e r
j
. (31)
7
Considering H

as a perturbation, the change in energy of the ground state can be calculated by


standard perturbation theory. The rst order contribution vanishes. The second order contribution
writes:
E
(2)
=
6e
4
r
6

k=0

l=0
[k [

i
r
i
.n [ 0)[
2
[l [

j
r
j
.n [ 0)[
2
E
1k
E
10
+ E
2k
E
20
. (32)
In this equation, k (l) labels the excited states of the rst (second) atom, [0) is the ground state of
the atoms (we avoided to put an indice recalling which atom is concerned, since there is no risk of
confusion) and n is the unit vector along the line joining the two nuclei (the direction is irrelevant).
The sums run, in principle, over all the excited states, but in practice only on those which are
connected to the ground state, by o-diagonal matrix elements of the dipole moments

d
1
=

r
i
or

d
2
=

r
j
. For atoms with J = 0 ground states, i.e. with spherical shapes, the sum in Eq. 32 is
limited to J = 1 states, but involves a summation over the magnetic quantum numbers. Using the
Wigner-Eckart theorem to relate the matrix elements of dierent magnetic quantum numbers, it is
easy to see that that the quantity E
(2)
does not depend upon the orientation of the vector n, as
intuitively expected.
The physical meaning of Eq. (32) is rather clear. Classically, two neutral objects with spherical
symmetry, even if they are locally charged have no Coulomb interactions. All their multipole moments
are vanishing. Quantum mechanically, spherical neutral atoms, have zero electric dipole moments only
on the average. They are uctuating. There is a non vanishing probability for having the two atoms
with non zero dipole moments and therefore experiencing a Coulomb interaction. The van der Waals
force is thus a purely quantum force originating from quantum uctuations.
It may be of interest to make two remarks. First, Eq. (32) is not valid if r is of the order of the size
of the atoms. At short distances, the interaction should be repulsive, due to the Pauli principle: the
latter forbids to put simply electrons at the top of each other; this is only possible if the electrons of
one atom are put at unoccupied orbits of the other, which requires a strong increase of the kinetic
energy. The repulsive nature of the atom-atom interaction at short distance is often embodied by
Lennard-Jones potentials. Second, it may be worthwhile to notice that the expression in Eq. (32) is
almost but not exactly proportional to the product of the electric polarisabilities of the atoms. The
electric polarisability is dened as the the ratio between the induced electric dipole acquired by an
atom in a static electric eld (considered as uniform for simplicity) and the magnitude of this electric
eld. It is given in second order by
=

k=0
[k [

i
er
i
.n [ 0)[
2
E
k
E
0
. (33)
For a matter made of these atoms, the dielectric constant is given, in the dilute limit, as
= 1 + 4n
at
, (34)
where n
at
is the density of atoms.
8
4.3 The van der Waals force and retardation eects
When he was working at the Philips company in Eindhoven, Casimir got interested into the behaviour
of the van der Waals interaction at large distances. Two colleagues of him, Verwey and Overbeek,
were studying experimentally colloidal suspensions. It seems that a simple model based on the van
der Waals force successfully reproduced their observations, but failed for dilute suspensions. They
interpreted their observation as due to a weakening of the van der Waals force at large distance [911].
They were thinking that retardation eects were the cause of this weakening. If the van der Waals
interaction is interpreted as due to the interaction between uctuating dipoles, the uctuation of one
dipole takes some time before inuencing the other atom when the interdistance is large enough and
vice-versa. They approached Casimir and asked him whether he could calculate this eect. A little
bit later, the answer was given in paper by Casimir and Polder [12]. We are not going to enter in the
details of this complicated calculation, but we will give the general ideas and the results.
In Eq. (32), only the static (Coulomb) interactions are introduced. How to cope with retardation
eects in Quantum Mechanics? Retardation eects are linked with the perturbations of the radiation
eld which propagate at nite speed. One has thus to introduce this radiation eld. This is usually
done by introducing a time-dependent vector potential in the hamiltonian. Adopting the Coulomb
gauge and the minimum substitution principle, this is equivalent to replace the momentum of the
electron p
i
by p
i
+
e
c

A(r
i
, t). With such a prescription, one has to add, along with H, a second
perturbation of the form:
H

=
e
mc

A(r
i
, t). p
i
+
e
2
2mc
2

A
2
(r
i
, t). (35)
The procedure is fairly standard. The system of the two atoms with a relative distance d is enclosed
in a cubic box with perfectly conducting walls. The vector potential is written as an expansion on
the normal modes

A(r, t) =

s
(a

k,s
e
it
+ a

k,s
e
it
)

k,s
(r), (36)
where

X

k,s
is the vector eld characteristic of the mode

k, s, duly normalized. The operators a and
a

are the usual destruction and creation operators. The total hamiltonian is written as:
H = H
1
+ H
2
+ H
rad
+ H

+ H

= H
0
+ H

+ H

. (37)
where H
rad
is the free radiation hamiltonian. The change in the ground state energy of the total
system, due to the perturbation H

and H

, is then calculated to second order. Note that to be


consistent, the calculation should include second order terms in the rst part of H

(linear in a and
a

) and rst order terms in the second part of this operator as it is already quadratic in a and a

.
Finally, the size of the box is extended to innity, while keeping the distance d xed. Needless to
9
say that the calculation is rather cumbersome. For detail, we refer to the original paper. We simply
quote the results.
For small r (actually smaller than the absolute value of the non-diagonal matrix elements of the
dipole operator), the London result (Eq. 32) is recovered. For large r (in principle larger than the
above-mentioned quantities), the following simple results is obtained:
E
(2)
=
23c
4r
7

2
. (38)
The van der Waals interaction is weakening as the distance increases and factorises in the polaris-
abilites of the atoms.
In the same paper, Casimir and Polder investigated also the interaction of an atom with a conducting
wall. The principle is the same: put the atom in the box at a xed distance d from a wall and let
the size of the box become innite while keeping a wall xed. In this case, the hamiltonian H

is the
Coulomb interaction between the atom and the wall, which is taken as the interaction of the dipole
moment of the atom and its image. The dipole moment is then considered as the corresponding
quantum operator. Note that the Coulomb atom-wall is neglected in the case of two atoms, since the
walls are eventually removed to innity. It is interesting to quote the results. For small distances, the
change of energy is given by:
E
(2)
atomwall
=
3
8d
3

k
[k [

i
er
i
.n [ 0)[
2
, (39)
where n is the unit vector perpendicular to the plane. For large distances, one has
E
(2)
atomwall
=
c
8
3
1
d
4
, (40)
where
1
is the polarisability of the atom.
Casimir was intrigued by the simplicity of the results, especially the one of Eq. 38 and was wandering
whether they can be more general. After all, these results were derived using the standard apparatus
of perturbation theory (to second order in the ne-structure constant , see later). He once discussed
these results with Niels Bohr, who is said to have replied [13]: Why dont you calculate the eect by
evaluating the dierence of zero point energies in the electromagnetic eld?. Of course this requires
to calculate the normal modes in the presence of the atoms, which is very hard. Casimir realized that
the calculation could be more easily performed for the case of the cavity as it is done in Section 2
and published his result in Ref. [14].
10
5 The nature of the Casimir force
5.1 Introduction
Although the existence and magnitude of the Casimir eect is now well established, there is still some
controversy concerning its nature and its interpretation. The Casimir eect looks universal. Formula
(27) indeed solely depends upon the the constants and c and upon the interdistance d. It is therefore
considered as a property of the electromagnetic vacuum, modied by the presence of the condensor
plates. On the other hand, it is tempting to interpret the Casimir eect as a generalized van der
Waals interaction between two gigantic molecules, the conducting planes. In this perspective, the
Casimir eect is reduced to an ordinary (though quantal) electromagnetic eect. It is then surprising
that this eect is not dependent upon the ne structure constant . Actually, it can be shown that the
independence upon results from the implicit hypothesis of perfectly conducting planes. When this
hypothesis is released, correcting terms in should be added. The result (27) appears to be correct in
the the limit of very large values of . In the following we will give simple arguments supporting this
assertion. We will also discuss the relation between the Casimir eect and the quantum uctuations
of the electromagnetic vacuum. Since this question is still under debate, we will limit ourselves to
general considerations.
5.2 The dependence of the Casimir eect on the ne structure constant
Actual metals are not perfectly conducting. They are characterized basically by two quantities: the
plasma frequency
pl
and the skin depth . For frequencies above
pl
, the conductivity basically
goes to zero. The quantity measures the distance up to which electromagnetic waves penetrate the
metal. A perfect conductor is characterized by innite
pl
and = 0. In actual metals,
pl
and 1/
depend upon the ne structure constant and vanish when 0. We turn to the simplest model
for real metals, namely the Drude model, to describe qualitatively what is happening. We closely
follow here Ref. [15]. Basically, in the Drude model, electrons are moving independently under the
inuence of the electric eld and they are subject to a friction force. Let E = E
0
e
it
be the applied
electric eld. The Newton equation of motion for the electron can be written as:
m
e
d
2
x
dt
2
= eE
0
e
it

dx
dt
. (41)
where m
e
is the electron mass and is the friction parameter. The solution is an oscillatory function
of time x(t) = x
0
e
it
, with
x
0
=
eE
0
m
e
( + i)
, (42)
11
where we have introduce the reduced friction parameter = /m
e
. It is then easy to calculate the
induced current (j = endx/dt). The result gives readily the conductivity ( = j/E) under the form
=
e
2
n
m
e
1
i
, (43)
where n is the electron density. The plasma frequency is given by

2
pl
=
4e
2
n
m
e
. (44)
The skin depth, which is dened by

2
=
2[[
c
2
. (45)
in general, becomes in the Drude model
=
c
_
1
2

2
pl

2
+
2
_1
2
. (46)
In this model, indenitely increasing
pl
automatically implies 0. In practice, typical frequencies
of interest are larger than , and becomes

c

pl

2
. (47)
The frequencies that are relevant for the Casimir eect are those with a frequency smaller than c/d.
The perfect conductor approximation requires therefore that c/d
pl
. Combining this relation with
Eq. 44 gives the following condition:

m
e
c
4nd
2
. (48)
For typical cases (copper plates separated by a micrometer), the rhs is of the order of 10
5
. Condition
(48) is comfortably satised by the physical value of . The standard Casimir result can then be
regarded as the limit of the true result which is dependent on the nature of the metal. For
large , one expects corrections to the Casimir result which could be put in series of negative powers
of . This result may be obtained very roughly by saying that for real metals, the limits of the cavity
become somewhat transparent to the electromagnetic eld and that the eective width of the cavity
becomes d + 2. One thus expects, instead of the relation (27)
E
S

c
2
720(d + 2)
3

c
2
720d
3
_
1
6
d
+ ...
_
. (49)
12
In this equation, the dots indicate higher order powers in 2/d. Owing to Eqs. 47,44, it is easily seen
that the latter ratio is proportional to 1/

. The corrections thus disappear as . It is also


interesting to verify that the second term in the parenthesis of Eq. 49 becomes negligible compared
to unity when condition (48) is fullled.
It is also interesting to look at the 0 limit. This limit is a little bit tricky as the typical size of
atoms, the Bohr radius
2
/m
e
e
2
, scales as 1/. Therefore, n scales as
3
,
pl
scales as
2
and goes
as 1/
2
. At very low , the plates become transparent to the radiation and the Casimir eect goes
away as 0. At low , the Casimir eect is expected to be put in a series of increasing positive
powers of .
As all ordinary electromagnetic eects, the Casimir eect goes away when the ne structure constant
goes to zero. The distinctive feature of the Casimir eect is that it reaches a nite value as .
5.3 The Casimir eect: vacuum property or interaction between neutral objects?
The Casimir eect is often pointed as an evidence of the reality of quantum uctuations of elds in
vacuum. Just to quote a typical example, Weinberg in his introduction of the cosmological constant
problem states [16]:
Perhaps surprinsingly, it was a long time before particle physicists began seriously to worry about
[quantum zero point uctuation contribution to ] despite the demonstration in the Casimir eect
of the reality of zero-point energies.
There are more recent quotations of this type. In his book Particle Astrophysics, Perkins says [17]:
That this concept [the vacuum energy] is not a gment of the physicists imagination was already
demonstrated many years ago, when Casimir predicted that by modifying the boundary condi-
tions on the vacuum state, the change in the vacuum energy would lead to a measurable force,
subsequently detected and measured by...
This kind of statements should be appraised after a close examination of the meaning of the expres-
sions reality and quantum uctuations. Let us start with the second one. The vacuum can be
considered as the quantum ground state of a eld, say the electromagnetic eld in the case of our
discussion. There is little doubt that there are quantum uctuations of observables associated to this
eld in the ground state, as it is for any observable which does not commute with the hamiltonian.
The simplest observables are the electric and magnetic elds themselves. The expression quantum
uctuations is also used to denote the zero point energy of the vacuum which is interpreted as the
energy generated by the quantum uctuations. Had the electromagnetic eld been vanishing with
certainty, would it be natural to expect a vanishing energy. When one relates the Casimir eect to
the quantum uctuations, it is to the second meaning of these words that one refers.
When boundaries are imposed to the electromagnetic eld, the latter is changed. The question arises
whether the ground state (the vacuum) of the electromagnetic eld is changed. Physicists have been
13
reluctant for a long time to admit that the energy of the vacuum is changed (advocating that the
zero point energy is innite and thus that its physical meaning is suspicious). The experimental
measurements of the Casimir eect have given support to the idea that the zero point energy is
perhaps unphysical, because it cannot be measured directly, but its variations when the geometry is
changed are physical since they are observed. Before discussing this point, let us mention that nobody
questions the change in the uctating properties of the electromagnetic eld when boundaries are
introduced. We will come to this question later.
Let us examine the reality of the change in the zero point energy as revealed by the Casimir eect.
The experiments are realized with condenser plates which, even in the limit of perfect conductors,
are not merely a mathematical device serving to conne the electromagnetic eld in a restricted
region of space. They are composed of atoms or molecules which interact with the electromagnetic
eld. The force which is measured is in fact the force between the material plates. In some sense,
it can be viewed as a van der Waals interaction between two gigantic molecules. It is the point of
view adopted by many physicists, who consider that the usual calculation, based on the change in

, is heuristic [15]. In other words, it is an accident that it gives the expression of the force
between two conductors. A similar example is provided by the energy of a smooth charge in classical
electrostatics which is given by W = 1/2
_
d
3
r
_
d
3
r

(r)(r

)/[r r

[ or, also, by the energy of the


electric eld W = 1/(8)
_
d
3
r[

E(r)[
2
. This second expression cannot be viewed as an evidence of
the reality of the electric eld but as an alternative expression of the self-interaction of the charges
which, heuristically, gives the correct magnitude of this self-interaction energy. The reality of the eld
and its extension outside of the sources cannot be proven by the action on a test charge as it is often
stated in elementary courses. By looking at the eect of a test charge, one probes the interaction
between the original source and the test charge. Bringing a test charge changes the nature of the
problem. The reality of the electromagnetic eld exists, but is revealed by other kinds of phenomena
like the Hertz experiment or the pair creation in a Coulomb eld.
Let us now examine the relation between the van der Waals eect and the Casimir eect at the light
of uctuations of the vacuum, in the sense of uctuations of operators in the vacuum. The van der
Waals interaction between two atoms is the change of the energy in the system of the two atoms
due to presence of the Coulomb interaction, which couples to the uctuating dipoles of the atoms
(this is translated as the interaction to the second-order perturbation term). The Casimir-Polder
interaction is the change of the energy of the system made of the two atoms and the ground state
of the electromagnetic eld, due to the presence of the Coulomb interaction, which couples to the
uctuating dipoles of the atoms as before, but also due to the interaction between the electrons and
the electromagnetic eld, which couples to the uctuations of the latter in the vacuum. When one of
the atoms is replaced by a conducting plate, the uctuations of this atom somehow disappear, but
the interaction is given by the change in energy of the whole system (atom, wall and electromagnetic
eld), due to the same causes. The change in dimension of the system is reected through the change
in the power in the dependence of the eect upon the distance. When the second atom is replaced
by the second plate, the interaction is solely due to the interaction of the plates with the uctuating
electromagnetic eld.
It is interesting to note that the interaction between two conducting plates can be constructed without
reference to zero point energy. First, we want to mention the method invented by Lifshitz [18,19] to
14
calculate the interaction between dielectrics. The starting point is the quantum uctuations of the
electromagnetic eld in large bodies. It is rst argued that the uctuations of the eld are linked to
the uctuations of the polarisation density, that the uctuations at dierent points (understood as
involving scale larger than the atomic size) are uncorrelated and that the mean square of the elds at
any given point is xed by the change of energy implied by the appearance of a dielectric constant.
The Maxwell stress tensor can be calculated and forces may be derived by taking derivatives. The
formalism is cumbersome and has not been published totally (which has hampered interest in this
kind of approach). Let us just quote some results. Lifshitz obtained an explicit expression for the
force between two innite dielectric bodies of dielectric constant with plane surfaces facing each
other at a distance d (expression (90.1), p. 369 of Ref. [20]). From this expression, several limits
can be obtained. The limit 1 corresponds to the dilute limit and the interaction between two
molecules can be obtained. The perfect conductor limit is obtained as (this corresponds to
the vanishing of the electric eld inside the bodies). All the results of Section 3 and the Casimir
formula can be obtained this way.
Let us also mention that the Quantum Field Theory can be formulated without any reference to
zero point energy. For instance, the Casimir eect may be calculated (perturbatively) in terms of
Feynmann diagrams with external legs, i.e. in terms of S-matrix elements making no reference to the
vacuum. We refer to Ref. [15] for more information.
In conclusion there is a strong division between physicists regarding the interpretation of the Casimir
eect. For many of them, the latter is a manisfestation of the vacuum energy. For many others, the
Casimir eect is the interaction between two large polarisable bodies. It does not tell upon the
quantum uctuations more than any one-loop eect in quantum electrodynamics, like the vacuum
polarisation of the Lamb shift.
6 The Casimir eect in Cosmology
It has been suggested that the Casimir eect or rather the vacuum energy could account for dark
energy. The rst suggestion dates to Einstein who introduced a cosmological constant in his funda-
mental equations of general relativity coupling the structure of space-time and the energy content:
R


1
2
g

R = 8G(T

+ E
v
g

). (50)
Einstein introduced the last term in the rhs, or equivalently g

, with the cosmological constant


= 8GE
v
, in order to manage a static solution. After the discovery of the expansion of the
Universe, he dropped this term, that he considered as the greatest mistake in [his] life. Later, with
the development of quantum eld theory, the possibility of the existence of a vacuum energy was
taken seriously (it seems that Zeldovich was the rst person to formulate such a possibility [21]) ,
especially after the experimental verication of the Casimir eect. It was even taken more seriously
when it was realized that the expansion of the Universe seems to be accelerating [22,23]. Recent
observations require a value of = (2.140.1310
3
eV )
4
at the present time [24]. The name dark
15
energy has been dubbed for this energy. The zero point energy of the electromagnetic eld has been
proposed as a candidate for the dark energy. We have given serious reservations on the reality of the
zero point energy, but this suggestion meets diculty beyond these reservations. First of all, the zero
point energy density is in principle innite. Of course, one may admit that the contribution of the
very high fequencies should be cut somewhere, rendering the energy density nite. The natural cut
should come from gravity and can be taken as the Planck scale. One then has

v
= c
k<kcut
_
d
3

k
(2)
3
k = ck
4
cut
. (51)
If k
cut
is taken as the inverse of Planck length
PL
= 1.2 10
19
GeV, one obtains a vacuum energy
of the order of 10
121
GeV fm
3
. This should be compared to the critical energy density
c
5 GeV
fm
3
and the part of about 75% taken by the dark energy. The supposed electromagnetic vacuum
energy is thus enormously too large and there is no clear mechanism to reduce it. Furthermore, one
should add in principle the contribution of the vacuum energy for the other fundamental elds. This
has led to the crisis of the cosmological constant and to a serious questioning about the reality of
the vacuum energy of the elds. Furthermore, there are plenty of condensates in the standard model
which also contribute to dark energy in principle. The conclusion is that the (perturbative) vacuum
energy (of the elds) has probably no real physical meaning or at least that our understanding of its
properties, especially concerning its coupling to gravity, has to be claried.
7 Conclusion
This short review aimed at presenting the main current ideas about the Casimir eect, its interpreta-
tion and its relevance to Cosmology. It does not reect the increasing activity linked with Casimir-like
eects in the nanoworld. See Ref. [25] for an introduction.
8 Appendix. Regularisation of Eq.23
As an example, we chose the regularisation provided by substitution (28) or
f(x) =

_
x
2
dt

te
t
. (52)
The rst three derivatives are given by
f

(x) = 2x
2
e
x
2
, (53)
16
f

(x) = (4x + 4x
3
)e
x
2
(54)
and
f

(x) = (4 + 20x
2
8
2
x
4
)e
x
2
, (55)
respectively. All derivatives at innity are vanishing. It is also easy to verify that f

(x) vanishes at
x = 0 and that f

(0) = 4 and that higher order derivatives at x = 0 are either vanishing or are
proportional to positive powers of . The rst term in the curly braket of Eq. (14) being equal to
f(0), formula (27) results directly from the application of the Euler-McLaurin formula to the function
(52), at the limit of vanishing .
17
References
[1] C. Itzykson and J. B. Zuber, Quantum Field Theory, McGraw-Hill, New York, (1985)ch.3.
[2] T. J. Torres, http://web.mit.edu/ttorres/Public/TheCasimirEect.pdf
[3] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions, Dover, New York, (1970)
886.
[4] M. J. Sparnaay, Physica 24 (1958) 751.
[5] B. V. Derjaguin, N. V. Churaev and V. M. Muller, Surface Forces, Plenum, New York, (1987) 886.
[6] S. K. Lamoreaux, Phys. Rev. Lett. 78 (1997) 5.
[7] T. Ederth, Phys. Rev. A 62 (2000) 062104.
[8] F. London, Trans. Faraday Soc. 33 (1937) 19.
[9] E. J. W. Verwey and J. T. G. Overbeek, Trans. Faraday Soc. 42B (1946) 117.
[10] E. J. W. Verwey, J. Phys. and Colloid Chem. 51 (1947) 631.
[11] E. J. W. Verwey, J. T. G. Overbeek and K. van Nes, Theory of the Stability of Lyophobic Colloids,
Elsevier, Amsterdam, (1948).
[12] H. B. G. Casimir and D. Polder, Phys. Rev. 73 (1948) 360.
[13] P. W. Milonni, The Quantum Vacuum, Academic Press, New York, (1994).
[14] H. B. G. Casimir, Kon. Ned. Akad. Wetensch. Proc. 51 (1948) 793.
[15] R. L. Jae, Phys. Rev. D72 (2005) 021301.
[16] S. Weinberg, Rev. Mod. Phys. 61 (1989) 1.
[17] D. Perkins, Particle Astrophysics, Oxford University Press, Oxford, (2003).
[18] E. M. Lifshitz, Zhurnal eksperimentalno

i i teoretichesko

i ziki 29 (1955) 9
[19] E. M. Lifshitz, Soviet Journal JETP 2 (1956) 73.
[20] L. D. Landau and E. M. Lifshitz, Electrodynamics of continuous media, Pergamon Press, London,
(1960).
[21] Y. B. Zeldovich, JETP Lett. 6 (1967) 316.
[22] N. Straumann, [arXiv:astro-ph/0203330].
[23] N. Straumann, Dark Energy, Lect. Notes Phys.721 (2007) 327.
[24] M. Tegmark et al [SDSS Collaboration], Phys. Rev. D69 (2004) 103501 [arXiv: astro-ph/0310723].
[25] P. W. Milonni, http://www.phys.nyu.edu/LarrySpruch/Milonni.pdf
18